Mobile devices (e.g., mobile phones) are good platforms for running deep learning applications, e.g., object or speech recognition, since they can record video and audio, and take pictures.
A simple implementation for such applications would send the content (e.g., video or images) back to central servers, which requires internet connection and consumes mobile data. Considering that the recognition task is cheap (cheaper than training tasks), we can implement mobile apps to do the recognition tasks on the mobile devices. In this way, we can run the applications off-line, which is very convenient for users.
This JIRA issue is going to re-implement some SINGA functions on mobile app development platforms to create a mobile app (android/ios) for object/speech recognition.